Estimates of variable stepsize Runge-Kutta methods for sectorial evolution equations with nonsmooth data

DOI10.1016/S0168-9274(01)00126-XMaRDI QIDQ1612427zbMATH OpenOpenAlexFDO

Authors Barnabas M. Garay, Wolf-Jürgen Beyn

Publication date 22 August 2002

Published in Applied Numerical Mathematics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/s0168-9274(01)00126-x

Banach space error bounds initial value problem nonsmooth data Runge-Kutta methods sectorial operator equations semilinear evolution equations stability variable stepsize

Mathematics Subject Classification ID

Abstract parabolic equations (35K90) Numerical solutions to equations with nonlinear operators (65J15) Numerical methods for initial value problems involving ordinary differential equations (65L05) Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations (65L06) Mesh generation, refinement, and adaptive methods for ordinary differential equations (65L50) Error bounds for numerical methods for ordinary differential equations (65L70) Error bounds for initial value and initial-boundary value problems involving PDEs (65M15) Method of lines for initial value and initial-boundary value problems involving PDEs (65M20)

Review text

The purpose of the paper is to derive error estimates for variable stepsize Runge-Kutta methods as applied to semilinear evolution equations \[ \dot{u}= Au + f(u),\quad u(0)=u_0\in X^\alpha, \] where \(A\) is a sectorial operator in some Banach space \(X\) which satisfies standard resolvent conditions, \(X^\alpha\), \(\alpha\geq 0\), denote the classical fractional order spaces derived from \(X\), and the nonlinearity \(f: X^\alpha\mapsto X\) is assumed to be globally Lipschitz. The hypotheses made on \(A\) and \(f\) are the following: (A0) \(A\) is a densely defined closed linear operator on \(X\) satisfying the spectral condition \[ \sigma(A)\subset \Sigma_{a,\theta}=\{z\in{\mathbb C}\setminus\{a\}:\;\theta < |\arg(z-a)|\leq \pi\} \] for some real \(a,\theta\) with \(\theta\in (\frac{1}{2}\pi,\pi)\) and the resolvent estimate \[ |(zI-A)^{-1}|\leq \frac{M}{1+|z-a|}\quad \text{whenever} z\not\in\Sigma_{a,\theta} \] for some positive constant \(M\). (F0) For a fixed \(\alpha\in [0,1)\) a mapping \(f:X^\alpha\mapsto X\) is given which satisfies for some \(L\geq 0\) the Lipschitz condition \[ |f(u)-f(\tilde{u})|\leq L|u-\tilde{u}|_\alpha\quad \text{for all} u,\tilde{u}\in X^\alpha. \] The main theorem is the following: Assume that conditions (A0) and (F0) are satisfied and that the Runge-Kutta method is strictly \(A(\theta^c)\)-stable (being \(\theta^c=\pi-\theta\)) and of order \({\mathcal P}\). In addition, assume that the stability function of the Runge-Kutta method, say \(r(z)\), satisfies \(r(\infty)=0\). Let \(T>0\) be arbitrary. Then there exist constants \({\mathcal K}>0\) and \(h_0>0\) such that for all finite stepsize sequences \((h_1,\dots,h_N)\), \(N\geq 1\), with \(H=\sum_{n=1}^N h_n\leq T\) and \(h_\ast=\max_{1\leq n\leq N} h_n\leq h_0\) we have the estimate \[ |U_N - \bar{u}(H)|_\alpha \leq {\mathcal K} \left( \left(\frac{h_\ast}{H}\right)^{\mathcal P} + \frac{h_\ast}{H^\alpha}\log\frac{H}{h_\ast}\right) \] where \(U_N\) stands for the variable stepsize Runge-Kutta solution at time \(H\) and \(\bar{u}(t)\) denotes the solution of the initial value problem \[ \dot{u}= Au + f(u),\quad u(0)=u_0\in X^\alpha. \] The constants \({\mathcal K}\) and \(h_0\) depend on the data \(T, \alpha, |u_0|_\alpha, |f(u_0)|, \theta, a\), on the constants \(M, L\) from (A0), (F0) as well as on the stability function of the Runge-Kutta method.

Recommendations

Cites work

Cited in

(13)

This page was built for publication: Estimates of variable stepsize Runge-Kutta methods for sectorial evolution equations with nonsmooth data

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1612427)